[Numpy-discussion] Float and locale formatting fixes merged in trunk

[Charles R Harris]

On Sat, Feb 14, 2009 at 10:34 AM, David Cournapeau <cournape at gmail.com>wrote:

> On Sun, Feb 15, 2009 at 1:28 AM, David Cournapeau <cournape at gmail.com>
> wrote:
>
> >
> > Note that this is windows x64, whose support has never been on part
> > with 32 bits ones. I will look at those.
>
> I looked at the problem, and I have no clue about what's going on. To
> make things interesting, the free 64 bits compilers are horribly buggy
> (both V 14 and V 15), and actually segfault or give bogus parsing
> errors when they encounter a problem. Of course, it builds fine with
> the mingw-w64 compiler...
>

My work around fixed this problem. I think David's configuration test fails
for this function but it is really there, so when we define the function
there is an error. The workaround was giving the numpy function a different
name, npy_tanhf, then #define tanhf npy_tanhf.  Since all the tanhf
references are replaced before the compilation pass the problem goes away.
Ideally, the presence of tanhf should be detected, so I guess that's where
the official fix should be. All the numpy tests were passing on the
Windows_64 platform.

Chuck
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20090214/1b0d6b92/attachment.html>